Dependence of higher-order correlations and information compression on temporal resolution in neuronal data
Studying the higher order interactions in complex interacting systems based on limited data is a challenging task. Traditional methods rely on pre-defined assumptions on the connectivity structure of the underlying generative models. In this paper, the inference of the underlying structure in complex systems given a limited binary dataset is done by a novel approach ( Minimally complex models ) with minimal prior assumptions. Our results demonstrate that orders of interaction and the number of components required for encoding of information vary with the choice of temporal resolution that the data is studied or recorded in. This procedure accounts for optimal time resolution in which the presence of the higher order interaction is accompanied by maximal information compression. The model provides systematic framework for inferring interaction's orders and accurately comparing the significance of these type of interactions across different co-variates.